Who Needs Emotions? The Brain Meets the Robot

obot emotion 301
regime is marked with positive valence and balanced arousal, contributing
to a contented affective state. The understimulated regime (large positive
values) is marked with negative valence and low arousal, contributing to a
bored affective state that can eventually decline to sorrow. The overstimulated
regime (large negative values) is marked with negative valence and high
arousal, contributing to an affective state of distress. Another factor is progress
toward achieving the desired goal of the active behavior. Success in achieving
a goal promotes joy and is tagged with positive valence. Prolonged delay
in achieving a goal results in frustration and is tagged with negative valence
and withdrawn stance. It is also possible for the active emotion to either
contribute to or inhibit the activation of other emotions, making it difficult
for a creature to be both happy and angry simultaneously, for instance.
Because there are potentially many different kinds of factor that modulate
the robot’s affective state (e.g., behaviors, motivations, perceptions), this
tagging process converts the myriad factors into a common currency that
can be combined to determine the net affective state. For Kismet, the A, V,
S trio is the currency the emotive system uses to determine which response
should be active. In the current implementation, the values of the affective
tags for the releasers are specified by the designer. These may be fixed constants
or linearly varying quantities.
Emotion Elicitors and Arbitration
All somatically marked inputs (e.g., releasers, the state of each drive, etc.)
are passed to the emotion elicitors. There is an elicitor associated with each
basic emotion “gateway” process (e.g., anger, fear, disgust). The elicitor determines
the relevance of its emotive response based on the myriad factors
contributing to it. In a living creature, this might include neural factors, sensorimotor
factors, motivational factors, and cognitive factors (Izard, 1993).
Each elicitor computes the relevance of its affiliated emotion process and
contributes to its activation. Each elicitor can thus be modeled as a process
that computes its activation energy, E emot(i), for emotion, i, according to the
following function:
excite
inhibit
E ()= i R ()+ i Dr ()+ i Em ()− i Em ()+ i Bh () i
emot emot emot emot
where R emot(i) is the weighted contribution of the active releasers, Dr emot(i)
excite is the weighted contribution of the active drive, Em ()
emot i is the weighted
contribution of the other active emotions that excite this process, Em i
emot
emot
emot
inhibit ()
is the weighted contribution of the other active emotions that inhibit this
process, and Bh emot(i) is the weighted contribution of the behavioral progress
toward the current goal.